Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
This skill enables agents to recognize and capture the successful completion of non-trivial tasks—the golden path—and preserve them as permanent skills. It is designed for automation builders who need agents to remember project-specific facts, such as deployment steps or credential locations, without repeated human explanation.
By proactively harvesting workflows after debugging sessions or complex operational tasks, the agent converts temporary session knowledge into a structured library of reusable capabilities.
WHO IT'S FOR
AI agent / automation builders
implement meta-learning capabilities in coding agents
Platform / DevEx teams
standardize project-specific operational workflows for AI agents
AI coding assistant power users
reduce repetitive prompting for complex project tasks
Technical leads / architects
codify tribal knowledge into machine-readable agent skills
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Follow the documented setup, then try a first task.
I’m using
npx project install with agent auto-detection
Installs: Installs the self-learning-skills repository bundle for agents detected in the current project context. · Claude Code, Cursor, Codex, Cline, OpenCode, other agents detected by the skills CLI
Before you start
The npx command must already be available.
A supported agent must be installed so the skills CLI can detect it.
Run this command in a terminal to install the repository bundle into detected agents.
npx skills add kulaxyz/self-learning-skills # this project (auto-detects agents)
Installs: Installs the self-learning-skills repository bundle globally for detected agents across projects. · Claude Code, Cursor, Codex, Cline, OpenCode, other agents detected by the skills CLI
Before you start
The npx command must already be available.
A supported agent must be installed so the skills CLI can detect it.
Run this command in a terminal to install the repository bundle globally.
npx skills add kulaxyz/self-learning-skills -g # global — all your projects
Installs: Installs the repository’s Cursor rule adapter; harvested rules subsequently go under .cursor/rules/learned/. · Cursor
Before you start
Git must already be installed.
A terminal with the documented mkdir and cp commands must be available.
Cursor must already be installed.
Run from: Target Cursor project root; the clone command creates the referenced self-learning-skills subdirectory inside it, while .cursor/rules is created for that project.
From the target project root, clone the repository in a terminal.
Installs: Appends the repository’s generic cross-tool self-learning instructions to the target project’s AGENTS.md file. · Codex, Zed, Aider, Gemini CLI, other agents that read AGENTS.md
Before you start
The curl command must already be available.
The target agent must read AGENTS.md.
Run from: Target project root containing, or intended to contain, the AGENTS.md file.
From the target project root, run this terminal command to append the generic adapter to AGENTS.md.
A multi-step workflow completed during the current session
Evidence of a passing check, such as a successful test, clean exit, green build, or reproduced fix
At least one failed approach that was tried and ruled out
Use the self-learning skill to capture the proven workflow we just completed as a project-local reusable Agent Skill. Base it on the workflow and validation evidence from this session, identify the failure pattern it avoids, include at least one concrete dead end we ruled out, and do not record secret values—record only where authorized users can find them.
Claude
Codex
Cursor
Gemini CLI
OpenCode
Cline